Environment model-based system testing of real-time embedded systems

The behavior of real-time embedded systems (RTES) is driven by their environment. Independent system test teams normally focus on black-box testing as they have typically no easy access to precise design information. Black-box testing in this context is mostly about selecting test scenarios that are more likely to lead to unsafe situations in the environment. Our Model-Based Testing (MBT) methodology explicitly models key properties of the environment, its interactions with the RTES, and potentially unsafe situations triggered by failures of the RTES under test. Though environment modeling is not new, we propose a precise methodology fitting our specific purpose, based on a language that is familiar to software testers, that is the UML and its extensions, as opposed to technologies geared towards simulating natural phenomena. Furthermore, in our context, simulation should only be concerned with what is visible to the RTES under test. Our methodology, focused on black-box MBT, was assessed on two industrial case studies. We show how the models are used to fully automate black-box testing using search-based test case generation techniques and the generation of code simulating the environment.